2
\$\begingroup\$

Normally when I design suppression filters I use π-filters for data lines and T-filters for supply lines as supplies generally have a lower output impedance than data pins.

I now have the situation where a long cable carries both data and supply and it picks up (radiated) interference.

But is the output impedance of the supply relevant when suppressing this radiated interference? Should I keep π-filters on the data lines and T-filters on the supply lines? Or can I better switch to all π-filters (assuming the interference behaves as if coming from a high-impedance source).

The data cable carries a 5 V supply and 4 single ended UART data signals (115k2). These go into/come from a 'black box' for which I only have basic documentation. I don't know what's inside, I just have to work with it. I can't change anything on these signals I'm afraid. As mentioned, all data lines are fitted with π-filters and the supply line is fitted with a T-filter.

The π-filters (currently) consist of 2x 100 pF and a 300 Ω resistor. The T-filter consists of 2x ferrite bead (300 Ω @ 100 MHz) and 100 nF. These are of course to be further fine tuned.

The PCB contains a bridge sensing circuit with ADC. Even without modulation, the ADC gives out a different/wrong reading when the RF carrier is present. Experimenting with ferrite clamps on the supply/data cable and the cable connecting the bridge shows that both cables participate in picking up the noise and somehow transferring it to the ADC.

I tested for immunity at 10 V/m from 80 MHz to 1 GHz. Only for specific ranges of frequency does the ADC start to give out wrong readings. E.g. around 84 MHz, 205 MHZ and 500 MHz.

While researching the problems I experience I was wondering whether the impedance of radiated interference should be related to the circuit impedance or more to a 'general' (probably unknown) impedance.

\$\endgroup\$
11
  • \$\begingroup\$ The best choice cannot be generalized; it is application specific and, even more than that, it depends whether using differential signalling or single-ended signalling. In other words, the devil is in the detail. \$\endgroup\$
    – Andy aka
    Apr 17 at 10:46
  • \$\begingroup\$ In addition to what Andy said in his comment, I'd like to say that in Wurth Electronics' TRILOGY OF MAGNETICS it's recommended to use T-type filter for low or unknown source impedance and low or unknown load impedance, and Pi-type filter for high or unknown source impedance and high or unknown load impedance. So, apparently, it doesn't matter how you treat the long-cable source because both of them can be used in any case. \$\endgroup\$
    – Rohat Kılıç
    Apr 17 at 10:58
  • \$\begingroup\$ Might be a good opportunity to move to fiber-optic cabling. \$\endgroup\$
    – rdtsc
    Apr 17 at 11:36
  • \$\begingroup\$ @RohatKılıç More to say, in the absence of any information (which is to say, question as it stands), either type is equally good (it might perfectly solve the OP's problem) and equally bad (it might completely destroy the OP's communications and increase radiation by peaking/ringing). What it's not saying, is that it can be applied safely without any ill effect -- this isn't like dusting bypass caps around a supply; here, the good and ill effects will most likely be strongly defined. \$\endgroup\$
    – Tim Williams
    Apr 17 at 12:49
  • \$\begingroup\$ To the OP -- a simple "yes" suffices for your first question, whereas I think you were looking for a more interesting answer. Perhaps you can add some information about what you are doing (what mode is the communication: protocol, speed, modulation, balanced, etc.; what is the supply: switching frequency and amplitude, DC voltage range, possible connected sources, etc.; what is the environment: emissions and immunity test levels, radiated and impulse, etc.) so a specific answer can be given. As you can expect from this comment, a truly general answer would fill a book or two. \$\endgroup\$
    – Tim Williams
    Apr 17 at 12:56

1 Answer 1

Reset to default
1
\$\begingroup\$

Yes.
By which I mean, place both, and then you can use DNE/DNP or 0 Ohm resistors to pick and choose until you get good results. Trying to figure out what's going to work for EMI is sometimes guess work, particularly if you can't perfectly qualify the problem. Give yourself options so you can experiment and find the best solution.
Don't fixate on filtering as well, think about your other options. Can the cabling be improved? Can you change the data signaling to improve noise rejection? Can you use ferrites to suppress the noise? How does the noise spectrum compare to your data Bandwidth? Can you increase the data power to raise it above the noise?

\$\endgroup\$
1
  • 1
    \$\begingroup\$ Placeholders on a PCB often don't cost that much compared to having to remake your PWBs and redo EMI testing (especially if you are renting an EMI lab for the testing). \$\endgroup\$
    – user4574
    Apr 18 at 15:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.